Redesigning conditionality

Pergamon

WorldDe~~elc~pmer~t. Vol. 25, No. 9, pp. 1399-1407. 1997 0 1997 Elsevier Science Ltd

All rights reserved. Printed in Great Britain 0305-750x/97 $17.00+ 0.00

PII: s0305-750x(97)00053-3

Redesigning Conditionality

PAUL COLLIER Centre for the Study of African Economies, Oxford, I/. K.

PATRICK GUILLAUMONT CERDI, University of Auvergne, France

SYLVIANE GUILLAUMONT CERDI, University of Auvergne. France

and

JAN WILLEM GUNNING* Free Universiq, Amsterdam, The Netherlands

Summary. - Conditionality has been used to induce reform. The rather disappointing results of this approach have led to a gradual redesign of the aid contract, toward a shorter period and more detailed conditions. The logical conclusion of this approach is for reforms to be priced piecemeal. Such an approach is fundamentally incompatible with government ownership of policy and hence reduces the credibility of reforms. Further. this attempt to induce reform has crowded out other uses of conditionality. We propose an alternative basis for aid allocation in terms of a retrospective assessment of a few major outcomes such as growth. We show how outcome measures can control for influences on growth over which the government has no control. 0 1997 Elsevier Science Ltd

Key MY&~ - Africa, aid, structural adjustment, conditionality

I. INTRODUCTION

Almost invariably, aid to developing countries is characterized by conditionality. Donors have tried to increase the effectiveness of their aid by imposing conditions, both on the way the aid money is spent and on government policies. Increasingly. particularly in the case of structural adjustment lending, they have attempted to induce recipient governments to reform policies by making program aid conditional on policy reform. In recent years conditionality has become more prominent, partly as a result of the shift from project to program aid, but also as a result of the debate on the effectiveness of aid. Donor governments, disappointed by the lack of success of aid particularly in Africa, actual or perceived, have chosen to impose more (and more stringent) conditions, often under pressure from their domestic aid constituencies. In many African countries this has led to a dominant donor presence, numerous donor representatives permanently based in ministries to assess whether agreed conditions are

being adhered to. The Stabex programme of the European Union offers a striking example of the present prominence of conditionality. Designed to generate automatic aid flows to compensate ACP countries which suffered shortfalls in the value of their commodity exports to the Community, it has been transformed into another source of structural adjustment lending.

The usual critique of conditionality has focused on the nature of the conditions rather than on the principle of conditionality (e.g.. Killick. 1984. Comia et al., 1987, Mosley. 1987). Donors have been criticized for the perceived negative consequences of fiscal conditions for social expenditure. There is. however, also a literature criticizing the

*We have benefited from discussions with Elliot Berg. Richard Blackburst, Chris Elbers, Ravi Kanbur. Michiel Keyzer, Bernard Petit, Wouter Tims and Howard White and from the comments of two anonymous referees. We would like to thank Anke Hoeffler for research assistance. Final revision accepted: April 1 I, 1997.

1399

1400 WORLD DEVELOPMENT

principle of conditionality. Some authors (e.g., Rodrik, 1989) have argued that aid in support of trade liberalization can be detrimental for the credibility of the government and may encourage policy reversal. This argument however, has been less prominent than the critique of the conditions themselves.

In this article we argue that donors should switch from attempting to “purchase” a prespecified menu of policy changes, to the allocation of aid on the basis of periodic overall assessments of government achievements. The structure of the paper is as follows. In Section 2 we discuss the possible objectives of conditionality. It is commonly assumed that the sole objective of conditionality is to induce policy change. We argue that there are several better objectives and that these all happen to conflict with the policy change objective. In Section 3 we argue that the present tranche-based (“short-leash”) lending generates three problems one of which is insuperable, and show how the alternative of periodic overall assessment can be operationalized. Section 4 concludes.

2. CONDITIONALITY: OBJECTIVES AND EFFECTIVENESS

Aid without strings attached is rare; donors typically attach conditions to their aid. The very presence of these conditions suggests that aid is not entirely a consensual relationship. There are various rationales for aid conditionality. We distinguish five objectives.

First, the donor may offer aid as an incentive for the recipient government to change its policies. This is the essence of program lending but also applies in traditional project lending where frequently conditions on policies are included: a power project often stipulates conditions on electricity tariffs, a project financing school buildings may involve conditions on school fees. This is the objective of inducement: the aim is to induce the government to do something it would not have chosen to do without the offer of aid.

Second, the donor may treat policies as exogenous but wish to concentrate aid in good policy environments. The rationale for selectivity between countries is that aid is more productive in a good policy environment. Indeed, the motivation behind the introduction of structural adjustment lending was that project aid was perceived to be ineffective in poor policy environments. Whereas the inducement rationale for aid seeks to make project aid effective by improving the policy environment, the alternative is to introduce selectivity which confines aid to countries which already provide adequate environments.

Third, the donor may want to ensure that the aid is spent on particular goods or services: on machines of a favoured design, equipment bought in the donor country, or food aid for a group considered deserving by the donor. In almost all such cases the donor attempts to restrict the way the aid may be spent in order to improve its effectiveness in raising the recipients’ welfare.’ This rationale may be called paternalism. Paternalism signals a disagreement between the donor and the government in the use of aid. Paternalism may be both undesirable and infeasible. Regarding desirability, the empirical basis for donor priorities is often questionable. For example, previous World Bank work on the benefits of primary education has exaggerated the social rate of return (Bennell, 1994, Appleton ef al., 1996). By contrast, recent World Bank studies on Ugandan road projects show an average social rate of return of 40% (Collier and Pradhan, 1996). Yet the donors have been keen to shift Ugandan public expenditure from roads to primary education. Such a shift is not necessarily wrong, but there is no evidence for the confident assertion that it is right. Similarly, donors insist that aid recipients should increase tax effort (or “resource mobilization” in the euphemistic language of the World Bank). A corollary is that aid cannot be used for reduced taxation. Most expenditures (in Africa typically about 80%) are private as opposed to public, and it is unlikely that the best use of aid is for 100% of it to be spent by the public sector. Reduced taxation is the most efficient form of transfer to private agents available to donors because of the costs of tax collection. Public choice theory suggests that governments will tend to over-tax relative to the social optimum because bureaucracies have an interest in their own expansion so that it is not even clear that the direction of donor pressure is appropriate. Because of the decay of the private formal sector, the tax base is narrow and so taxation may become dysfunctional at low shares of revenue to GDP. Even when desirable, paternalism may be infeasible. Fungibility implies that the donor’s control over the use of aid is largely illusory, regardless of conditions and monitoring.

Fourth, aid may be used as a mechanism of commitment (“lock-in”) for government policies. In that case the objective of conditionality is to provide a credible threat. A government committed to a policy reform, say trade liberalization, may want to protect itself and its successors against pressures for reversing the reform. It can achieve lock-in of the reform by agreeing to an aid package conditional on the reform being sustained. For this form of conditionality to be effective the cessation of aid in the case of policy reversal must, obviously, be perceived as a credible threat. In this case the donor acts as an agency of restraint, preventing undesirable policy changes by attaching unacceptable penalties

REDESIGNING CONDITIONALITY 1401

to them. Therefore, in this case the objective of conditionality is restraint. Although restraint de- pends upon the donor providing the government with an incentive, through conditionality, not to change a policy, and so may appear to be a form of negative inducement, the difference is that whereas inducement implies that the donor and the recipient do not have the same objectives, with restraint there is no policy disagreement.

Fifth, the objective of conditionality may be signding. Because information on government performance is costly, private agents may find it preferable to free-ride on the decisions of donors who are known to have invested considerable resources in monitoring performance. Donors may wish so to reduce private sector decision costs in order to stimulate investment. Potentially, agreement to donor conditions may be regarded as a signal of a prospective improvement in policy.’

At present the first of these five objectives is dominant, however, it has had only limited success. The link from aid flows to policy change may have been quite weak (see e.g., Guillaumont and Guillau- mont, 1994, 1995). The inducement effect of aid has persuaded governments to enter into commitments which they subsequently break, thereby weakening their reputation. For example, in spite of the importance attached by donors to trade reform, donor-supported reforms are not usually sustained. Oyejide et al. (1997) show that in 10 African countries which undertook trade liberalizations, there were reversals in seven countries. In many cases, for example in Kenya, there was a whole series of reversals, liberalization episodes being separated by periods in which trade controls were reintroduced or tightened. Killick (1996, pp. 213-214) gives a disturbing summary for the structural adjustment lending of the two Bretton Woods institutions:

As at April 1993, only five out of a total of 26 ESAF programmes had been completed within their planned period and eight had apparently broken down altogether. Three-quarters of World Bank adjustment loans had instalment tranche releases delayed because of non- implementation of policy conditions in 1980-88, the latest period for which data are publicly available.... The Bank’s Adjustmenf in Africa.. judges that only six out of 29 adjusting countries had achieved decisive improve- ments in macroeconomic policies (much the most important of which, Nigeria, shortly afterwards jumped off this pedestal)...

Indeed, the study to which Killick refers shows that those countries which had increases in net resource transfers were considerably less likely to improve their policies than those which experienced reductions.3 Further, even where aid has induced policy reform it may have done so in a context in which the reforms have insufficient credibility to

generate private investment. One reason for this is that inducement is liable to be time inconsistent. To the extent that the aid is temporary there may come a moment where the government no longer has an incentive to maintain the policy reform. Because private agents can recognize this danger, the receipt of aid can contaminate the credibility even of governments genuinely committed to reform. A second reason is that if the donor attempts to “buy” reform, the government has an incentive to max- imize its “price”. That is, the government has an incentive to exaggerate the social and political costs of reform. A corollary of the donor “buying” reform is that ownership is vested in the donor rather than the government. An extreme example of this was the statement by President Moi of Kenya in July 1995 that unless the donors desisted from complaints about civil rights he would reverse the economic reform programme. Such a lack of ownership of policy reform by the government is now understood to have contributed to the low credibility of African reforms and hence to have impaired supply response (World Bank, 1994).

Further, the objective of inducement conflicts with the other four objectives of conditionality. Inducement conflicts with selectivity. If the objective is selectivity aid should obviously be concentrated on those countries with the best policy environments. This offers however, the least scope for policy improvement so that if the objective is inducement aid should be concentrated on those countries which currently have unsatisfactory policy environments. The choice between the two objectives should be determined by the elasticity of policy with respect to aid. Given that policy is evidently less elastic with respect to aid than donors had originally hoped, the implication is to shift from inducement to selectivity.

Inducement conflicts with paternalism. If the objective is to induce the government to undertake reforms which it does not otherwise want to do. the most effective aid is unrestricted: the inducement is more potent if it can be used for what the government wants rather than for what the donor wants. This conflict often arises in practice: structural adjustment loans have been characterized by increasingly wide-ranging conditions on how the aid can be spent.

Inducement also conflicts with restraint. The essential feature of the restraint role is that the government chooses to subject itself to an agency whose long term objectives are the same as its own, but which is not susceptible to the pressures which may otherwise require the government to deviate from these objectives. The government will only place itself in such a position vis-ri-vis the agency if it is confident that the long-term objectives are indeed the same as its own. Hence, the donor role is


that of Ulysses’ crew members, tying him to the mast at his own request. Like Ulysses the government willingly gives up power in order to achieve its own objectives. Intrinsic to inducement, however, is a signal from the donor that its preferences diverge from those of the government.

Finally, inducement conflicts with signalling. The private sector becomes unable to distinguish between those governments which genuinely want to reform and those which are only doing so because of the inducement, the implication being that the former are more secure. Inducement reduces the information content both of aid and of government actions. The combination of reform with aid may erroneously be interpreted as time inconsistent.

Because inducement conflicts with the other four objectives of conditionality there is a tradeoff between them. Although this tradeoff has probably not been recognized, donors have assigned most weight to inducement. Since inducement has evidently been ineffective, it is important to assess its opportunity cost in terms of the other three objectives foregone.

The opportunity cost of inducement in terms of reduced selectivity is arguably important. If indeed policy is insensitive to aid but the productivity of aid is highly sensitive to policy then unless aid is concentrated in the better policy environments it will be wasted because its efficacy is overestimated. Greater selectivity need not imply that the countries which do not receive aid are written off. There is some evidence (Easterly and Levine, 1995) that the growth performance of individual African countries is influenced by that of their neighbors. Hence, an increase in the productivity of aid would benefit even those countries not receiving it as a result of these spill-over effects.

Now consider the opportunity cost of using aid for inducement in terms of its weakening of the restraint role. African governments are very short of commitment mechanisms for a whole range of policies to which private investment is sensitive. They have not used the General Agreement on Tariffs and Trade (GATT) (now the World Trade Organization WTO) to lock into trade liberalization.4 Outside the Franc Zone they cannot lock into convertibility, and their central banks are insufficiently independent to provide credible fiscal and monetary discipline. Investment rates in Africa are very low and it is now widely recognized that the main reason for this is policy uncertainty. Lock-in mechanisms are therefore important for promoting economic growth in Africa. Aid is not very well-suited for achieving lock-in but for the next decade it is likely to be one of the few mechanisms available, so that the weakening of the restraint role of aid implied by its use for inducement is costly.

The weakening of the signalling role of aid

implied by its use for inducement is also costly. Although African governments are already rated by commercial agencies there are two serious limitations to these ratings. First, they are concerned with debt valuations which reflect past histories of indebtedness rather than with current economic performance. Second, recent work (Haque et al., 1996) shows that when ratings are related to country characteristics, a significant African dummy re- mains. That is, African countries are rated as more risky than observed characteristics seem to warrant. Hence, private agents are currently making choices on the basis of very poor information.

Concern over how donor funds are spent is clearly legitimate. Since fungibility undermines the capacity of donors to achieve any specific expenditure objectives, however, there is little point in patemal- ism as an objective. Hence, the use of aid for inducement does not have an opportunity cost in terms of the objectives of paternalism.

3. TRANCHE-BASED LENDING VERSUS “PERIODIC ASSESSMENT”

The past ineffectiveness of inducement as the objective of conditionality has led to the gradual redesign of the conditionality contract. Aid released on the promise of policy change backed by the threat of aid secession has come to be regarded as ineffective. The solution has been partly to replace promised-based aid with performance-based aid, and partly to shorten the period for which the contract applies. Aid is no longer released on the basis of promises of reform over an extended future period but in tranches on the implementation of each element of an agreed sequenced program. Even the first tranche, which was formerly released upon signature of a statement of intent. is now commonly released upon completion of policy change. Subse- quent tranches are conditioned upon a specific sequence of reforms, the failure to implement any part of this sequence in principle suspends the entire program, a structure commonly termed “short-leash lending.”

Short-leash lending has not, however, overcome the problem of limited donor credibility. First, by conditioning the entire gross aid inflow upon each of multiple conditions. the penalty is frequently incommensurate with the triggering policy failure. This is somewhat analogous to the limited credibility of a nuclear deterrent in international relations. Second, the suspension of aid would usually trigger a macroeconomic crisis leading to a default on debt service. Hence, the lender would suffer from inflicting the penalty and so has an incentive to avoid imposing it. This produces the phenomenon of defensive lending.

REDESIGNING CONDITIONALITY 1 -to3

The most likely donor reaction to the problem of incommensurate penalties is the development of a “flexible response” (again, much as the solution devised in international relations). Each proposed policy reform would be “priced”: that is program aid would be disaggregated. Failure to implement any particular reform would then not hold up the entire aid disbursement but only that part of it regarded as commensurate with the policy failure. The penalty becomes proportional to the offence: breach of a minor condition gives rise to only a small aid reduction. One implication of such an approach is that donors would abandon the attempt to specify the sequencing of policy change. Aid agreements would become priced menus of policy changes from which the government chose.

Such piecemeal “pricing” of reforms does not solve the problem of defensive lending. The government may still choose not to “sell” enough reform in order to service its debt, seeking rescheduling instead. The ability to service the debt becomes dependent upon the pace of reform rather than its level. This problem will become more acute as a result of democratization. Around the world. governments follow a political business cycle: difficult policies are deferred during the approach to an election. Thus, democratic governments will indeed choose to reduce the sales of reforms in some years. Consider, for example, a government which has an above-average record both on policy reform and on performance. but which completely halts further reform in the approach to a difficult election. It would evidently be ridiculous, and therefore incredible, for donors to suspend aid under these circumstances. In practice, what would happen at present is that the donors would waive conditions and reschedule. thereby discrediting themselves.

To avoid this the servicing of existing debts needs to be subject to different conditions from those attached to net flows. This can be done by the donors either reducing the stock of debt. or by making commitments to service it subject to explicit performance conditions which are weaker than those governing net flows. Given the limited credibility of the donors, there is some advantage to economizing on promises and so tackling the debt problem by writing off part of the stock rather than committing themselves to servicing the flow. This is what donors are currently attempting in the debt reduction initiative for the highly indebted poor countries (HIPC). But, even this only changes the nature of the credibility problem faced by donors, since “one-off” debt forgiveness faces its own time-consistency problem.

Were short-leash lending so redesigned as to avoid the “nuclear deterrent” and defensive lending problems, it might prove effective in delivering short-run policy change. It would, however, face one

further and less tractable problem. namely that it exacerbates limited ownership. Short-leash aid is the “pricing” of reform carried to its logical extreme: the donor buys reforms and therefore owns them. As economic reforms proceed the impediment to private responses increasingly becomes the credibility of the policy environment. A policy environment which has been “purchased” by donors is not one about which private agents can be confident.

In order to signal ownership. governments need to have a “longer leash.” Governments need to be responsible for formulating policies as well as for implementing them. Performance would then be assessed more upon the attainment of objectives than upon the implementation of particular policy measures. In order to assess the attainment of objectives. performance would need to be evaluated over a longer period than the current tranche-based system. Whereas policy changes can be monitored against a timetable delineated in months, outcomes cannot be monitored for periods of less than a year and may require rather longer. As the review period is extended it becomes less credible since conditionality is ultimately a political process. For example, were the review period to be a decade. governments might reasonably doubt whether the assessment process would be maintained for such a long period. Some of the outcomes in which donors are most interested are slow changing. most notably poverty, and so conditioning aid upon these outcomes becomes problematic. Review periods can perhaps be extended to three years, and the monitoring of poverty outcomes can perhaps both be improved and be supplemented by the monitoring of policy Inputs. but the weight placed upon such poorly measured outcomes should unfortunately be correspondingly reduced relative to their importance.”

Outcomes must be evaluated with reference to some norm and donors must ensure that the norms which they employ are seen as legitimate. One way of achieving legitimacy is to define the norm in terms of average performance of a peer group. so that for example, in Africa governments would be evaluated relative to the African average. To the extent that the adoption of such a criterion for aid succeeds in improving performance, the mean of the indicator itself changes so that it becomes gradually more demanding. Note that such an improvement in performance need not depend upon incentive effects of aid (inducement). Simply by reallocating aid to environments in which it can best be used would raise mean performance, and greater selectivity would encourage what may be the most potent process of policy improvement, namely the need to emulate more successful neighbours (see Herb% 19901. The African average is likely to be politically acceptable: a government will find it difficult to argue that this average cannot be attained. Setting


mean African performance as the critical evaluation point is therefore less controversial than other and necessarily arbitrary levels. For non-African countries it should be possible to select a yardstick with a similar legitimacy.

The evaluation of performance is in most respects more difficult than the evaluation of policy change since policies are generally more observable than outcomes. Further, even when outcomes are fully observable, they are usually not fully under the control of the government. While judging by outcomes rather than policies gives the government both the freedom and the responsibility of determin- ing how best it will achieve those outcomes, it therefore risks punishing governments which are unfortunate and rewarding those which are fortunate. Effective evaluation should therefore both attempt to correct for circumstances outside the control of the government, and perhaps also continue to extract some information directly from the observation of policies. The donor may wish to evaluate several aspects of performance such as economic growth, poverty alleviation, child mortality and civil rights.6 Clearly, the extent to which it is possible to correct objectively for circumstances outside the govem- ment’s control will differ between such objectives. An element of judgement is unavoidable but there is scope for quantitative analysis, particularly in the case of economic growth. We consider this case as an example.

Growth can be measured directly through constant price GDP. In many countries the statistics are now produced on a timely basis and so the actual growth rate is a major indicator of performance. The issue is the extent to which it needs to be adjusted and supplemented. Since the end use of the indicator is to evaluate government performance, the actual growth rate should, to the extent that it is possible, be modified to make allowance for influences over which the government can have had no control but which are known to have affected the growth rate. Two factors beyond the control of the government might reasonably be allowed for, structural factors and temporary shocks. From the growth literature there are only two structural factors which appear significantly to affect the long-run growth rate and which have sufficient intuitive appeal that it should be possible to build a consensus that it would be proper to make allowance for them. The first of these is an aspect of location, namely. whether a country is landlocked. Sachs and Warner (1995) find that landlocked countries have grown significantly more

slowly over the past 30 years. This is empirically important for Africa: many African countries are landlocked.

The second structural factor to be allowed for is the ethno-linguistic diversity of the population. There is, fortunately, a standardized measure of

ethno-linguistic fragmentation which calculates the chances of any two citizens of a country being from the same group. This measure is available for nearly all countries showing their ethno-linguistic composition as of 1960. Easterly and Levine (1995) pioneered the use of this measure in growth regressions. They found that countries with a high degree of ethno-linguistic fragmentation grow significantly more slowly. Again it is not hard to imagine why this might be the case. Ethnic and linguistic divisions make it more difficult to reach a stable political consensus and so make long-term commitments, notably, investment, more risky. They also may reduce the basis for trust. therefore increasing transactions costs. While such effects are still controversial a donor may want to take into account a country’s “fractionalization” in evaluating its growth performance.

We illustrate by introducing location (LAND) and ethno-linguistic fractionalization (ELF) into a stan- dard growth regression. The other regressors are dummies for the 196Os, 1970s and 198Os, dummies for Africa and Latin America, the real per capita income level in 1960 (INCOME), the square of this variable, the log of the primary school enrollment ratio in 1960 (LMSCHL), the number of (actual or attempted) assassinations of politicians or govem- ment officials, as a proxy for political instability (ASSASS), and indicators of financial, trade and fiscal policies (DEPTH, the ratio of M2 to GDP; BLACK, the black market exchange rate premium; and the government deficit, DEFICIT). (see Table 1).

Table 2 reports the average growth rates for all sub-Saharan African countries and adjusts these rates for the fact that the countries differ as to location and

Table 1. Structural factors in developing counfry growth. 196&92”

Variable Coefficient t-ratio

DVM6O -0.06669 -0.708 DUM70 -0.06504 -0.690 DUMSO -0.07772 -0.830 AFRICA -0.01916 -4.226 L4 TINCA -0.01918 -4.654 INCOME 0.03436 1.471 INCOME2 -0.00287 -1.919 LNSCHL 0.01794 1.708 ASSASS 18.286 -2.811 DEPTH 0.01827 2.519 BLACK -0.01932 -5.240 DEFICIT -0.12270 -3.110 ETHNIC -000012 -2.287 LAND -0.00677 -2.142 No of 182 observarions: Adjusted R’: 0.546

aThe coefficients are corrected for heteroskedasticity. Data source: Penn World Tables and World Bank.

REDESIGNING CONDITIONALITY I405

Table 2. Adjusting African growth rates for structural differences

Unadjusted growth Adjusting for fragmentation Adjusting for location Growth adjustment

Angola 0.057 78.0 -0.160 0 0.217 Burundi 2.745 4.0 0.743 -0.416 2.418 Benin -0.047 62.0 0.035 0 -0.082 Botswana 7.159 51.0 0.169 -0.416 7.406 CAR -0.336 83.0 -0.22 1 -0.416 0.301 Gate d’lvoire 1.570 86.0 -0.258 0 1.828 Cameroon 1.722 89.0 -0.295 0 2.017 Congo 2.682 66.0 -0.014 0 2.696 Ethiopia 0.715 69.0 -0.05 0 0.765 Gabon 2.019 69.0 -0.05 0 2.070 Ghana -0.636 71.0 -0.075 0 -0.561 Gambia 2.174 73.0 -0.099 0 3.273 Guinea B -0.532 n.a n.a. 0 -0.532 Burkina Faso 1.578 68.0 -0.038 -0.416 2.033 Kenya 1.828 83.0 -0.221 0 2.049 Liberia -1.014 83.0 -0.22 1 0 -0.793 Lesotho 3.318 22.0 0.523 -0.416 3.21 I Madagascar -1.141 6.0 0.718 0 - 1.859 Mali 1.485 78.0 -0.160 -0.416 9.062 Mozambique -3.888 65.0 -0.000 0 -3.886 Mauritania 1.471 33.0 0.389 0 I .08?, Mauritius 3.271 58.0 0.084 0 3.18; Malawi 1.400 62.0 0.035 -0.416 1.781 Niger - 1.037 73.0 -0.099 -0.416 --0.531 Nigeria 0.793 87.0 -0.207 0 I.063 Rwanda -0.282 14.0 0.621 -0.416 -0.487 Sudan -0.140 73.0 -0.099 0 -0.041 Senegal 0.040 72.0 -0.087 0 0. IX’ Sierra Leone 1.025 77.0 -0.148 0 1.173 Somalia 0.25 1 8.0 0.694 0 -0.44:~ Swaziland 3.464 n.a. n.a. -0.416 3.X8 I Chad 1.030 69.0 -0.050 -0.416 -0.56.1 Togo 2.007 71.0 -0.07s 0 3.082 Tanzania 0.513 93.0 -0.343 0 0.857 Uganda ~ 1.232 90.0 -0.307 -0.416 - 0.508 South Africa 1.408 88.0 -0.282 0 1.690 Zaire -0.941 90.0 -0.307 -0.416 -0.3 1 x Zambia -0.750 82.0 -0.209 -0.416 -0.124 Zimbabwe 0.926 54.0 0.133 -0.416 1.211)

social fragmentation. The adjustment factor for ethnic diversity was obtained by subtracting the individual country’s ethno-linguistic fragmentation index from the average for all sub-Saharan African countries and multiplying with the coefficient for Ethnic in the regression. The adjustment factor for geographic location was obtained using the coefficient for Land from the regression and multiplying this coefficient with the average of the landlocked dummy for the region. The adjusted growth rate was obtained by correcting GDP growth for ethno- linguistic fractionalization and geographic location. For example Burkina Faso grew on average 1.578% per year. Taking into account that Burkina Faso’s ethnic diversity is higher than in the average African country and that Burkina Faso is landlocked, the country’s growth rate should have been 2.033%. Somalia on the other hand has low social fragmenta-

tion according to the index and is not landlocked, so Somalia’s growth rate of 0.25% becomes -0.443% after adjustment.’

Now consider how to allow for temporary shocks which affect the growth rate but which are unrelated to policy. The obvious candidate for this is the terms of trade variation (see Collier and Gunning with Associates, 1997). Terms of trade changes do not directly affect constant price GDP, because the latter concept already values output at a hypothetical set of prices rather than at actual prices. Indirectly however, terms of trade shocks alter both output and investment. To an extent, both the extent of terms of trade shocks and their consequences are endogenous to the policy environment. Anti-export bias tends to increase export concentration and so increase terms of trade volatility. During shocks. taxation and investment policies affect how well the economy


adjusts. Hence, there is a limit as to how much an evaluation should make allowance for terms of trade changes. Deaton and Miller (1996) use time-series data for 35 African countries to study the effect of changes in export prices on GDP and find that a terms of trade change worth 1% of national income on average has changed output in the year of the shock by 0.6% over-and-above any effects via investment. This provides a feasible GDP correction factor for terms of trade changes.

For example, applying this approach to Uganda in the mid-1990s would result in a premium over actual growth rate of 0.5% for being landlocked and being ethno-linguistically diverse. offset by a deduction of 2% because of the coffee boom.

While this example for GDP growth shows how measured performance can be corrected for factors not under control of the government (such as landlockedness or trade shocks) it is clear that for other outcome measures (e.g., poverty alleviation) data availability precludes a justification of quantitative targets in terms of a regression analysis. Inevitably, judgement will have to be exercised to determine the effect of luck (good or bad) on a country’s performance.

Outcomes are generally harder to monitor than policies. Social indicators are particularIy problematic. In many African countries health and education measures are collected irregularly and often unreliably (Srinivasan, 1994). Donor effort on improved data collection on outcomes is. however, useful in itself and would strengthen the incentive structure. Further, since the concern with policy reform has shifted from matters such as the exchange rate to the composition of public expenditure, policy change has become considerably more difficult to observe. There is in practice no way of enforcing donor conditions on the composition of public expenditure because audited public accounts are seldom available on a sufficiently timely basis for the implementation of penalties to be seen as legitimate. Some desired outcomes are practically or intrinsically difficult to observe so that donors will be driven back to observing the policies which are believed to generate them. By prioritizing information on outcomes, however, donors will gradually enhance the collection of the relevant data.

4. CONCLUSION

Donor conditionality is used for several objectives. Some of these are compatible with a consensual relationship between donor and recipient governments. For example, if the objective is restraint then conditionality helps the government to resist pressure for changes in policies which both it and the donors consider desirable. If the objective

is inducement, however, the relationship is not consensual. Structural adjustment lending has made inducement the dominant objective of conditionality. The rather unsatisfactory outcomes of this process have led to a gradual redesigning of the conditionality contract with the emergence of “short-leash” lending. The specification of policy change has become more detailed, and has been monitored over short periods corresponding to tranches. The logic of this approach is for reforms to be priced individually. While this redesign overcomes some of the dis- advantages of early adjustment lending, we have criticized it on two counts: it is inconsistent with government ownership, and it is at the cost of better uses of conditionality.

The attempt to buy policy changes actually exacerbates the problem of lack of ownership of policies on the part of the government. Without government ownership the persistence of reform may have insufficient credibility to induce a strong supply response. We have argued that a necessary condition for government ownership is that the design of policy should be done by the government. The normal process of government is one in which objectives are determined and the policy instruments appropriate for their attainment are then chosen. Present donor arrangements undermine this process because they specify the policies which the govem- ment must adopt in order to receive aid. If donors are not to specify policies they must evaluate govem- ment performance in some other manner. We have suggested that periodic evaluation of a few outcomes such as GDP growth, would be preferable. One disadvantage with switching from policies to outcomes is that it can reward good luck. To minimize this we have suggested correcting the growth rate both for structural features and temporary events over which the government has no control. Recent advances in the applied economics of growth and shocks have made such adjustments more reliable. Conversely, the policy reforms on which donors have focused have tended to become more difficult to observe with any accuracy. Hence, the advantage of greater observability which used to apply to policies is considerably diminished.

The pricing of reforms is the extreme form of the inducement objective of conditionality. We have argued that this is in conflict with four other objectives, selectivity, paternalism, restraint and signalling. The criteria for aid allocation which we propose are consistent with three of these. The allocation of aid on the basis of outcomes would concentrate aid in those countries in which it is likely to be most effective, thus meeting the objective of selectivity. By relinquishing the objective of inducement and hence the very public conflicts of priorities, it becomes possible for donors to be invited by governments to function as agencies of

REDESIGNING CONDITIONALIT\ 1407

restraint. Finally because aid would follow economic performing African economies some years earlier performance it would signal that performance. This than would otherwise be the case. may assist in attracting private capital to the best-

NOTES

I. Of course, donors also often impose restrictions for 5. Guillaumont and Guillaumont ( 1994) also advocate less unselfish reasons; the tying of aid is an obvious ex-post evaluations but envisage a shorter period. example.

2. Rodrik (1996) finds that at present donor behavior appears to be used by private agents as a perverse signal:

6, Such evaluations may require independent assess-

IMF agreements actually appear to reduce private capital ments. For example. the evaluation may involve surveys to

flows. This may reflect the low credibility with which measure changes in poverty conducted by an independent

private agents view conditionality as presently practiced. team, possibly jointly chosen by the donor and the recipient government.

3. World Bank ( 1994. Appendix Table 29).

4. This is signatory and Policy Review

changing. Zimbabwe. for example, is a 7. This approach of assessing performance on the basis has submitted itself to the GATT Trade of residuals in a cross-country regression is similar to that in process. Guillaumont et rd. ( 1988).

REFERENCES

Appleton. S., Hoddinott, J. and Knight, J. (1996) Primary Haque, N. U., Kumar, M. S.. Mark. N. and Mathieson. D. J. education as an input into post-primary education. (1996) The economic content of indicators of developing Oxford Bulletin qf‘ Economics and Sfufistics 58. 21 I- country creditworthiness. Mimeo. IMF Research Depart- 219. ment. Washington DC.

Bennell. P. (1994) Rates of return to education: Does the conventional pattern prevail in sub-Saharan Africa? Working Paper No. IO, Institute of Development Studies, University of Sussex. Brighton.

Collier, P. and Gunning J. W. with Associates ( 1997) Trade Shocks in Developing Countries. Clarendon Press. Oxford.

Herbst, J. (1990) War and the State in Africa. Intemarional Security 14(4), I 17-l 39.

Collier. P. and Pradhan. S. ( 1996) The economic recovery from Civil War. In Developing Uganda, ed. B. Hansen and M. Twaddle. James Currie, London.

Comia. A., Jolly, R. and Stewart, F. (1987) Adjustment wirh a Humun Face. Protecting the Vulnerable and Promot- ing Growrh. Clarendon Press, Oxford.

Deaton. A. and Miller, R. (1996) International commodity prices, macroeconomic performance and policies in sub- Saharan Africa. Journal of Africun Economie.r 5(AERC Supplement). 99-191.

Killick, T. ed. ( 1984) The Questfcrr Economic Stabilizozion: The IMF and the Third World. Heinemann. London.

Killick, T. (1996) Principals, agents and the limitations of BWI conditionality. The World Economy 16. 21 I-229.

Mosley. P. (1987) Conditionality as bargaining process: Structural adjustment lending 1980-86. E’ssa.vs in International Finance, No. 168. Princeton University Press, Princeton NJ.

Oyejide, A.. Ndulu. B. and Gunning, J. W. ! 1997) Regional integration and trade liberalization: Evidence from country case studies. introduction and overview. In Regional Integration and Trade Liberalization in Sub- Saharan Africa. Volume 2: County Case Studies, ed. A. Oyejide. B. Ndulu and J. W. Gunning. Macmillan. Basingstoke and London.

Easterly, W. and Levine. R. (1995) Africa’s growth tragedy: A retrospective, 1960-89. Policy Research Working Paper 1503. World Bank, Washington DC.

Guillaumont. P. and Guillaumont. S.. eds. (1994) Adjust- ment and Development. The E.rperience of the ACP Countries (Africa. Carabbean, PacijicJ. Economica, Paris.

Guillaumont, P. and Guillaumont, S. (1995) La conditio- nalitC B 1’Cpreuve des faits. In La Mgociation Commer- ciale et Financiere Intemationale. ed. M. Rainelli. Economica. Paris.

Rodrik. D. (19891 Credibility of trade reform: A policy- maker’s guide. The World Economy 12, 1-16.

Rodrik. D. (1996) Why is there multilateral lending’? In Annuul World Bunk Conference on Development Economics. 1995. ed. M. Bruno and B. Pleskovitc. pp. 167-l 93. World Bank. Washington DC.

Sachs. J. and Warner. T. ( 1995) Economic reform and the process of global integration. Brookings Papers on Economic Activifies 1, l-l 18.

Guillaumont, P.. Guillaumont, S. and Plane. P. (1988) Participating in African monetary unions: An alternative evaluation. World Development 16. 569-576.

Srinivasan. T. N. (1994) Data base for development analysis: An overview. Journal of De~~elopment Eco- nomics 44, 3-27.

World Bank (1994) Adjusrment in flfriccr: Reforms. Results und the Rood Ahead. Oxford Universitv Press. Oxford. i

Documents

Redesigning conditionality